Sasha (Augusto) Kielbowicz

Scientific and numerical software consulting

I help scientific and quantitative teams investigate slow or unreliable software, establish what is going wrong, and decide what to fix first.

My background combines a physics degree with engineering experience at J.P. Morgan, Qontigo, SimCorp, Mercado Libre, and AstraZeneca.

Discuss a diagnostic · Read my experience

Problems I can help investigate

• A numerical computation takes too long or costs too much to run.
• A research workflow is difficult to reproduce, test, or maintain.
• An AI workflow needs systematic evaluation of tool use, retrieval, or token consumption.

Start with one defined problem

Introductory conversation

We discuss the problem, its impact, and whether my experience fits. If it does, I propose a paid diagnostic with an agreed scope and fee.

Focused diagnostic

An investigation of one agreed computation or workflow, with the relevant inputs, access, and success criteria defined before work begins.

You receive:

• A reproducible baseline or an explicit account of what prevents one.
• Evidence-backed findings, including limitations and unresolved questions.
• Prioritized recommendations and a handover conversation.

The fee and calendar are agreed in advance. Implementation and ongoing support are separately scoped. A diagnostic may conclude that further optimization or development is not justified.

Discuss your problem

Selected work

Numerical performance — Qontigo

Problem: Convertible bond pricing calculations were too slow for the team’s real-time reporting needs.

Contribution: Investigated performance in the C# numerical library and implemented cache optimizations.

Outcome: Improved pricing performance to support real-time production risk calculations.

AI integration and evaluation — AstraZeneca

Problem: An R&D agent platform needed external integrations and a way to measure workflow quality.

Contribution: Built the external integration layer and automated evaluations for agent task execution, tool use, and retrieval.

Outcome: Delivered the integration layer to production and established repeatable evaluations for tracking platform quality.

Read the detailed experience

Further engineering support

After a diagnostic, I can separately scope numerical optimization, AI integration and evaluation work, research software engineering, or domain-specific language design.

For larger engagements, I work through Phorma, the consultancy I co-founded with Agustín Corbat (why I created Phorma).

I work remotely and in Buenos Aires on a limited number of defined engagements, with availability agreed in advance.

Resume / CV

Writing

ak.saxa.xyz/blog — technical posts on agentic AI, RSE, and scientific computing.

Contact

Discuss a problem (introductory conversation): calendar.app.google/txzedFUh1p7nvuAX8

LinkedIn: linkedin.com/in/augusto-kiel

GitHub:

• github.com/sashakile — scientific & numerical code
• github.com/akielbowicz — personal projects & talks
• github.com/charly-vibes — exploring the limits of Structured Vibecoding

“Namespaces are one honking great idea – let’s do more of those!” — Tim Peters, The Zen of Python